Our aim was to understand the role played by papG alleles in UTI pathogenesis, specifically their association with particular UTI syndromes, other VF genes and phylogenetic groups. Our findings demonstrate, in a stringently selected reproductive age women population by region, time and UTI syndrome, that papGII allele is strongly associated with pyelonephritis, with 68% of the isolates harboring this gene, whilst papGIII was more common in cystitis isolates, albeit at a lower level of influence, with a majority of 31% of the cystitis isolates harboring this allele only. These findings are in agreement with those of several other studies in different jurisdictions [9, 10, 31]. However, despite the stringent selection of the isolates for inclusion in the present study according to the uro-clinical category, it is possible that some misclassification of the uro-clinical syndrome might have occurred.

The predominance of the papGII gene in the studied pyelonephritis isolates, is in line with the demonstrated abundance of papGII iso-receptors in human kidney tissue, indicating an important role for this allele in ascending UTI infection as previously observed [32]. The papGII gene has been associated epidemiologically with pyelonephritis and urinary-source bacteremia in directed, usually PCR-based studies [33], and was demonstrated experimentally, with varying degrees of rigor, to contribute to kidney infection in murine and monkey models [34,35,36]. However, contrary to these findings, is the reported dominance of papG class II gene in pediatric cystitis isolates, albeit small number of isolates studied [37, 38], suggesting that associations of papG alleles with specific clinical syndromes may depend on the specific population studied, including age, gender, and geographical locale [37, 38]. It is thus important that comparisons of studies take into consideration VF gene carriage population dynamics.

On average, pyelonephritis and cystitis isolates, irrespective of papG allele gene status, more often belonged to phylogenetic group B2, and to a lesser extent group D, had higher VF gene scores, than fecal isolates which were mostly confined to groups A and B1. Our results are consistent with previous evidence of the domination of E. coli phylogenetic group B2 among extraintestinal pathogenic E. coli (EXPEC), and suggest that phylogenetic group B2 may be the main source, and hence presumably the original source of many VF genes in EXPEC [39]. These EXPEC VF genes are understood to be mainly inherited vertically within evolutionary ancestries, but can also be transferred horizontally between lineages, on gene bocks that contain multiple contiguous VF genes, commonly referred to as “pathogenicity-associated islands” (PAIs), or through plasmids [9, 39]. Although on average the UTI isolates had higher VF gene scores than the fecal isolates as previously observed [28, 40,41,42], the pyelonephritis isolates contained more individual VF genes, and consequently had higher aggregate VF gene scores, than the cystitis (or fecal) isolates, suggesting that VF gene repertoire plays a significant role in ascending UTI pathogenesis [43].

VF gene distribution by papG gene status revealed that papG positive pyelonephritis isolates on average contained more VF genes, and hence higher VF scores, than their papG negative counterparts, implying an association of papG with several other VF genes. This may be due to the fact that the pap operon can be located on chromosomal or PAIs which contain other VF genes, and hence are transmitted as a block of VF genes [6, 44]. The high VF gene carriage amongst the pyelonephritis isolates was mostly due to the contribution of papGII + isolates, which on average carried more VF genes than their papGIII counterparts as evidenced by higher VF scores, with 16 out of 18 VF genes analysed being much more abundant in the papGII isolates. This suggests an association of papGII with a wide variety of VF genes, compared to papGIII, and hence a possible increased inferred virulence potential of such strains, and increased capacity to cause ascending UTI. Likewise, the overall trend for cystitis isolates in relation to papG gene vs. other VF gene carriages, followed a similar pattern to pyelonephritis isolates. Taken together, these results strongly suggest that papG is an important VF gene in UTI pathogenesis, especially given that it is involved in the attachment of E. coli to the epithelial cells of the urogenital tract [45].

Interestingly, although papGIII gene was the most prevalent allele in cystitis isolates (31%), on average, papGII cystitis isolates had significantly higher VF gene scores (P = 0.03), suggesting increased inferred virulence potential for such strains as also demonstrated in the pyelonephritis cohort. This suggests an increased association of papGII with a wide range of VF genes, possibly due to their phylogenetic group B2 status as most of the papGII isolates derived from this phylogenetic group. However, even among the papGII isolates, pyelonephritis isolates had higher overall VF gene carriage than cystitis ones (data not shown), probably indicating presence of papGII clones within phylogenetic group B2, with differing levels of virulence, based on other factors, including possibly VF gene copy number, and other host and bacterial factors, as previously observed by us [41, 42]). Notably, predominance of the papGII allele among avian pathogenic E. coli isolates with high homology to human isolates has been previously reported [46], and also similarity between human and avian E. coli strains representing zoonotic potential has been demonstrated [47], suggesting that horizontal gene transfer of pathogenicity elements from chickens to humans may play a role in UTI pathogenesis [48].

Since we had a reasonable number of cystitis isolates that had a combination genotype of papGII and papGIII genes concurrently (n = 21), we analysed these isolates in relation to VF gene carriage. These isolates on average had the highest VF gene scores, with all VF genes tested, save for two (gafD and sfaS), having a prevalence of at least 67% each in these isolates. Hence, as per murine and mice studies [29], the inferred virulence capacity for such isolates was very high, and as such, we expected these strains to be confined to the pyelonephritis group, which was not the case. A further analysis of the patients from which the isolates originated revealed that they were all early UTI cases, suggesting that the bacterial strain might not have had enough time to cause ascending UTI. However, it is rather surprising that we did not find a high number of isolates containing both alleles amongst the pyelonephritis isolates.

We further analysed the association between the different papG gene alleles and other pap operon genes since they are encoded on the same pap operon. Among pyelonephritis isolates, papG negative isolates were more likely to be negative for all other pap genes tested (82%), suggesting complete absence of the pap operon or possible deletion [49], and hence inability to produce the P fimbria, or only positive for papC (8%), suggesting presence of an incomplete or truncated pap operon and hence incapacity to express P fimbria. This finding indicates that other VF adhesin genes or other bacterial or host factors may also be involved in ascending UTI pathogenesis, albeit on a smaller scale.

In the case of cystitis isolates, the majority (65%) of papG negative isolates, albeit lower proportion compared to the pyelonephritis isolates (82%), harbored no other pap genes, and hence were presumed to lack capacity to express P fimbriae. This finding suggests that for the pathogenesis of cystitis, other adhesins may be involved, including afa, sfaS, fimH, bmaE, gafD. Indeed, when these isolates were analysed for other adhesins, they, on average harbored a higher and wider variety of adhesins compared to their pap operon positive counterparts (data not shown). Many microorganisms have the genetic capacity to express different adhesins, providing access to multiple receptors and therefore increasing their pathogenicities [50]. Specifically, about 14% of our papG negative cystitis isolates were positive for all pap genes tested, suggesting possible deletion of the papG gene or presence of a yet to be described papG allele.

The majority of papGII positive pyelonephritis isolates (79%) were also positive for all the other pap operon genes tested, and hence had the ability to express P fimbriae under the right conditions [51]. In contrast, the pyelonephritis papGIII gene isolates demonstrated a limited capacity to express P fimbriae, as only 17% of the isolates presumably carried a complete pap operon, whilst the rest had incomplete operons with basically a uniform spread of other pap operon genes, mostly at 17% for each of the individual pap genes tested. This probably explains why papGIII isolates were more likely to be detected in cystitis isolates, as such isolates would have lacked the capacity to produce P fimbriae which is considered important in ascending UTI pathogenesis. It is not clear why the papGII gene was associated with a complete pap operon in pyelonephritis isolates, which calls for further studies to investigate this finding.

When compared by papG allele status, the majority of cystitis papGII + (46%) and papGIII + (38%) isolates contained all the other pap genes tested, and hence considered to have the capacity to express P fimbria. However, this was much lower than was the case with pyelonephritis isolates, where an overwhelming majority of the papGII strains had a presumed complete pap operon. These findings again seem to highlight that the capacity to express P fimbria is important in ascending UTI pathogenesis as observed by others [52, 53]. Furthermore, a large majority of isolates that contained both alleles concurrently (82%), contained a presumed intact pap operon. Although such isolates were, in the present study, confined to the cystitis isolates, we think that they had the capacity to cause ascending UTI over time as most of them were limited to early UTI cases.

Although the above argument that papGII operons from pyelonephritis isolates are more likely to be complete than papGII or papGIII operons from cystitis isolates, is plausible from a pathogenicity perspective, caution must be exercised in the interpretation of the findings. Firstly, the interpretations are based on the amplification of a limited region covering about 1 kb of the 9 kb pap operon, which limits the strength of the argument about the completeness of the pap operon. It is reasonable to argue that insertions or other recombination events are more likely to occur in the rest of the 8 kb region that was not investigated. Furthermore, some minor sequence variants have been described in the primer binding regions of papA and papE genes, and thus one can also hypothesize that such variants could have been present in the negative isolates. However, from an epidemiological viewpoint, we think that such minor variants would not have much effect on the overall picture.

Although our findings seem to suggest association of specific papG alleles with specific UTI syndromes, contrasting findings, such as the presence of strains with a combination of both papGII and papGIII genes amongst cystitis strains, and presence of a significant proportion of papGII isolates among cystitis strains, suggest that epidemiologic associations between individual papG alleles and specific clinical syndromes must be interpreted cautiously because such associations may be due primarily to other bacterial properties that are differentially linked with particular papG allele, possibly as part of a PAI rather than to any specific pathogenetic role of the papG allele itself. Like many other E. coli VF genes, the pap operon encoding P fimbria lies on PAIs [9, 54], which are large horizontally transferable genetic elements assumed to play an important role in the evolution of pathogenic E. coli [9, 54].

In 25% and 38% of pyelonephritis and cystitis isolates, respectively, both the major subunit gene (papA) and the distal adhesin gene (papG) were absent, suggesting presence of truncated P fimbrial operons as previously observed [55] or presence of minor variants. At least 95% of the papGII pyelonephritis isolates were positive for both fimH and other pap operon genes tested, and about 50% of them were positive for pap operon genes, fimH, and sfa/foc, implying capacity of these strains to express a wide variety of adhesins. This is in agreement with studies demonstrating that type 1, P, S, F1C, and Dr fimbriae, attach to different sites within the human kidney [32], and thus strains endowed with a diverse range of fimbrial types are more likely to have overall success during renal colonization [32]. Furthermore, P and type 1 fimbriae appear to act in synergy to promote colonization of kidney [56].

Although the papGI allele is relatively rare amongst clinical urinary isolates, it was surprisingly detected in 7% of the present isolates as follows; as the only papG allele in 1% of the cystitis isolates; in 3% each in combination with either papGII or papGIII, and finally as part of a concurrent combination with both papGII and papGIII genes in 1% of pyelonephritis isolates. To the best of our knowledge, this is one of the few, if any, studies in which the papGI allele was this much abundant, and may be attributed to differences in the distribution of papG alleles by geographical location, population cohort or other yet to be elucidated factors. The genetic make-up, or other aspects of the papGI + strains need to be more specifically determined to see if they belong to the same clonal group. Similarly, the papGIV gene, which has been rarely described in previous studies, was detected in only 1% of our isolates, and was only confined to the cystitis subgroup. More studies are needed to clarify the role of this papG allele in UTI pathogenesis.

The identification of a small proportion of cystitis isolates (3%) with the papGI and papGIII allele class combination highlights that this papG genotype, although rarely encountered among clinical isolates, is not limited to source strain J96 as was previously assumed [57, 58]. Previous reports indicate that this genotype is characteristic of a J96-like clonal group of E. coli strains of serotype O4:H5:F13 [55, 58], and some of these strains have caused cystitis, pyelonephritis, urosepsis, and bacteraemia of unknown origin, in parts of Europe and the United States of America (USA) [37, 38]. Both the papGI and papGIII alleles are associated with a papA molecule of the F13 serotype [55, 58], and may explain the present finding in which isolates that were papGI allele positive, also contained the papAH gene.

Strengths of this study include the large number of well-characterized cystitis, pyelonephritis, and fecal isolates, from the same geographical region and time period. This is important since human-associated E. coli strains can vary dramatically by region and over time [59, 60]. Other strengths include the extensive array of bacterial traits studied and the analysis of their distribution by phylogenetic group and uro-clinical syndrome.

Study limitations include the use of multiple comparisons, which can increase the chance of type 1 errors [61]. However, we regard our study as being of an exploratory nature rather than definitive, requiring future confirmatory studies, and is also designed to generate further hypotheses for future studies. Furthermore, virulence level was inferred based on molecular attributes, not in vivo assessment, and presence-absence testing for a defined set of VF genes can overlook other potentially important determinants of cystitis and pyelonephritis, including unrecognized VF genes [62], minor sequence variants of described VF genes [63, 64], or differences in VF gene expression [65]. Another limitation of the study is the possible overlap in the classification of isolates by uro-clinical syndrome into the 2 groups of cystitis and pyelonephritis. However, this is somehow compensated for by the large number of isolates studied. And finally, due to limited budget, whole genome sequencing was not performed, which could have shed some light into possible association of papG alleles with specific clones, presence of novel alleles, and whether the pap operons were truly disrupted.